Most tax returns selected at local service centers are chosen because they involve simple, readily identifiable problems usually removable by correspondence, or because they have a special feature such as an illegal deduction. Generally, the problems are identified by a computer. District offices select returns randomly sometimes for special research programs, but generally the returns are selected because they have good audit potential. The potential is discovered by a computerized system called the Discriminant Function System (DIF). In most cases, the decisionmaker is not the auditor. A weighted DIF measuring scale, developed by random sampling of returns, is used to determine the need for auditing. Returns included in the audit inventory are examined by classifiers, who determine the need for audit in light of appended material the computer could not see. The classifiers, then, either send returns to be audited to revenue agents who contact the taxpayer or to the taxpayer and a tax auditor. Under the DIF system, a person who overassesses his tax liability is less likely to be audited than the person who underassesses. Sometimes an error can be considered insignificant and the case closed, but there are no uniform criteria for deciding significance. IRS should revise the form letter it sends notifying taxpayers of unallowable items on their returns. Examiners should provide written explanations as to why they need the returns. The extent of the classifiers' influence should be measured. Taxpayers should be informed of right to appeal in all cases, especially the right to meet with the examiner's supervisor, before the examiner seeks agreement.